v0.0.8

Highlights

• High-level fmpose3d inference API with Hugging Face integration for easy model weight downloads.
• Extendable model registry and config dataclasses — scripts now use model_type instead of raw model paths.
• Animal pipeline support via DeepLabCut SuperAnimal, with batch-wise 2D estimation and robust status propagation.
• DeepLabCut is now an optional dependency (pip install "fmpose3d[animals]"), keeping the default install lightweight.

What's Changed

• Make DeepLabCut an optional dependency by @deruyter92 in #25
• Minor refactors for DeepLabCut integration by @deruyter92 in #23
• Inference API: add batch-wise image analysis for superanimal prepare_2d by @deruyter92 in #20
• Inference API: add handling for (partially) unsuccessful 2d estimations by @deruyter92 in #19
• Add huggingface functionality by @deruyter92 in #18
• Address timm deprecation warnings by @deruyter92 in #17
• Add animal pipeline support to FMPose3DInference (and add test suite) by @deruyter92 in #16
• Refactor model configuration with a model registry by @xiu-cs in #15
• Add high-level API for the pose3d inference pipeline by @deruyter92 in #14
• Add configs and model registry by @deruyter92 in #13
• Minor refactors by @deruyter92 in #12

Documentation and Developer Experience

• Updated README.md and animals/README.md with installation guidance for optional animal dependencies.
• Added and refreshed tests for config, inference API behavior, and edge cases.

Upgrade Notes

• If you rely on animal inference features, install the extra dependencies:
  • pip install "fmpose3d[animals]"
• Existing script integrations may need to switch from direct model path usage to model_type-based model selection.

Full Changelog: v0.0.7...v0.0.8